}

Status VerifySupportedSignature(TaggedValue signature) {
  if (signature.type() == TaggedValue::Type::TENSOR_SPEC) {
    return ::tensorflow::OkStatus();
  }
  if (signature.type() == TaggedValue::Type::TUPLE) {
    for (const auto& t : signature.tuple()) {
      if (t.type() != TaggedValue::Type::TENSOR_SPEC) {
        break;
      }
    }
    return ::tensorflow::OkStatus();
  }

    // Set up an arg per arg_shape with the specified type.
    for (int i = 0; i < arg_shapes.size(); ++i) {
      auto metadata_arg = metadata_proto.add_args();
      metadata_arg->set_kind(
          tensorflow::tpu::TPUCompileMetadataProto::Arg::PARAMETER);
      metadata_arg->set_dtype(dtype);
    }
    // Set up one dummy retval.
    metadata_proto.add_retvals();

  FunctionDef make_ref_float = FunctionDefHelper::Define(
      "RefFloatFn", {}, {"r:float"}, {},
      {{{"var"},
        "VariableV2",
        {},
        {{"dtype", DT_FLOAT}, {"shape", TensorShape({})}}},
       {{"r"}, "Identity", {"var"}, {{"T", DT_FLOAT}}}});
  *fdef_lib->add_function() = make_ref_float;
}

void AddRegularFunctionFunctionDef(FunctionDefLibrary* fdef_lib) {

  const tensorflow::Tensor& arg_tensor = cc_ctx->input(index);
  absl::Status cc_status;
  if (arg_tensor.dtype() != tensorflow::DT_RESOURCE) {
    cc_status = absl::InvalidArgumentError(
        absl::StrCat("Trying to obtain resource handle from Input[", index,
                     "], which is not type DT_RESOURCE."));
    status->status = cc_status;
    return nullptr;
  }
  const tensorflow::ResourceHandle& handle =

            metadata.type = TF_ATTR_INT;
          } else if (typestr == "list(float)") {
            metadata.type = TF_ATTR_FLOAT;
          } else if (typestr == "list(bool)") {
            metadata.type = TF_ATTR_BOOL;
          } else if (typestr == "list(type)") {
            metadata.type = TF_ATTR_TYPE;
          } else if (typestr == "list(shape)") {
            metadata.type = TF_ATTR_SHAPE;

  TF_Status* status = TF_NewStatus();
  // Create the variable handle.
  TFE_Op* op = TFE_NewOp(ctx, "VarHandleOp", status);
  if (TF_GetCode(status) != TF_OK) return nullptr;
  TFE_OpSetAttrType(op, "dtype", TF_FLOAT);
  TFE_OpSetAttrShape(op, "shape", {}, 0, status);
  TFE_OpSetAttrString(op, "container", "localhost", 0);
  TFE_OpSetAttrString(op, "shared_name", "", 0);
  if (!device_name.empty()) {

  }
  strings::StrAppend(&ret, "}");
  return ret;
}

string PrintTensor(const TensorProto& tensor_proto) {
  Tensor t(tensor_proto.dtype());
  CHECK(t.FromProto(tensor_proto));
  const int64_t num_elts = t.NumElements();
  switch (t.dtype()) {
    case DT_FLOAT:
      return PrintArray(num_elts, t.flat<float>().data());
    case DT_DOUBLE:
      return PrintArray(num_elts, t.flat<double>().data());

  }

  // Derive the output types. The result type is derived by using the
  // attributes attched to the result type of the signature. The attribute
  // value should be either in the attribute argument list or the derived
  // attribute from the input tensors. All the result type
  // are unranked, and shape inference should be applied afterwards.
  SmallVector<Type, 4> output_types;

  } else {
    return se::DeviceMemoryBase(const_cast<char*>(tensor.tensor_data().data()),
                                tensor.tensor_data().size());
  }
}

Status XlaTensor::AllocateShapedBuffer(DataType dtype,
                                       const xla::Shape& on_device_shape,
                                       xla::LocalClient* client,
                                       int device_ordinal) {
  xla::Shape on_host_shape =

  for (auto* n : arg_nodes) {
    int index;
    TF_RETURN_IF_ERROR(GetNodeAttr(n->attrs(), "index", &index));
    DataType dtype;
    TF_RETURN_IF_ERROR(GetNodeAttr(n->attrs(), "T", &dtype));
    (*recv_at_host_dtypes)[index] = dtype;
  }
  for (int i = 0, end = recv_at_host_dtypes->size(); i < end; i++) {
    if ((*recv_at_host_dtypes)[i] == DT_INVALID) {