size_t size = results.missing_unused_input_map_keys.size();
  tf_results->missing_unused_key_names.resize(size);
  tf_results->missing_unused_key_indexes.resize(size);

  for (int i = 0; i < size; ++i) {
    TensorId id = results.missing_unused_input_map_keys[i];
    tf_results->missing_unused_key_names_data.emplace_back(id.first);
    tf_results->missing_unused_key_names[i] =

                          << ") != terminator operands ("
                          << term->getNumOperands() << ")";
    return {};
  }
  // Verify number of tensors for inputs and outputs matches size
  // of the list in the signature def.
  if (input_names.size() != sig_def_inputs.size() ||
      output_names.size() != sig_def_outputs.size()) {
    main_op.emitWarning(

    return op.emitOpError() << "requires N to be at least 2, got " << op.getN();

  if (op.getShape().size() != op.getOffset().size())
    return op.emitOpError()
           << "requires sizes of shapes and offsets to be the same, got sizes "
           << op.getShape().size() << " and " << op.getOffset().size();

  auto ranked_dim =
      mlir::dyn_cast<RankedTensorType>(op.getConcatDim().getType());

  assert(input_shape.size() <= 32);  // Only 32-bit masks are supported.

  // Make sure ranges' ranks are consistent with the input.
  assert(input_shape.size() == begin.size());
  assert(input_shape.size() == end.size());
  assert(input_shape.size() == stride.size());

  for (int i = 0, e = input_shape.size(); i < e; ++i) {
    if (ShapedType::isDynamic(input_shape[i])) continue;

}

// Returns the number of elements in `range`.
template <typename Range>
size_t Size(Range&& range) {
  return range.size();
}

// Returns the total number of elements in a variadic number of `ranges`.
template <typename Range, typename... RangeTs>
size_t Size(Range&& range, RangeTs&&... ranges) {
  return range.size() + Size(std::forward<RangeTs>(ranges)...);
}

      }
      axis++;
    }
  }

  DenseIntElementsAttr size;
  if (matchPattern(op.getSize(), m_Constant(&size))) {
    int axis = 0;
    for (const auto& size_i : llvm::enumerate(size)) {
      if (size_i.value().getSExtValue() < -1) {
        return op.emitError(
            llvm::formatv("size[{0}] cannot be negative other than -1", axis));
      }
      axis++;
    }
  }

// all 0s and `size[i]` is equal to either -1 or `input.shape[i]`
// for each dim i, the output tensor is identical to `input`.
bool CanOptimizeIdentitySliceOp(Value input, Attribute begin, Attribute size) {
  // Checks if `begin` and `size` are i32 or i64.
  auto begin_attr = mlir::dyn_cast<DenseIntElementsAttr>(begin);
  auto size_attr = mlir::dyn_cast<DenseIntElementsAttr>(size);
  if (!begin_attr || !size_attr) {

    controls_to_idx.insert({control_and_idx.value(), control_and_idx.index()});

  if (controls_to_idx.size() != control_outputs.size())
    return errors::InvalidArgument("Control outputs must be unique");

  control_ret_nodes->resize(controls_to_idx.size());

  for (auto* node : GetOrderedNodes()) {
    auto it = controls_to_idx.find(node->name());

  if (!(verify_size(window_dimensions.size(), "window_dimensions") &&
        verify_size(window_strides.size(), "window strides") &&
        verify_size(padding.size(), "padding entries") &&
        verify_size(lhs_dilation.size(), "lhs dilation factors") &&
        verify_size(rhs_dilation.size(), "rhs dilation factors")))
    return std::nullopt;

  xla::Window window;
  for (size_t i = 0; i < window_dimensions.size(); i++) {