"tf.SomeOp2"() {device = "/device:CPU:0", _output_shapes = ["tfshape$"], f.Tin = [f32], f._read_only_resource_inputs = []} : () -> ()
  func.return
}

// CHECK-LABEL: func @basic
func.func @basic(
    %arg0: tensor<3x1xf32>,
    %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>>) -> (tensor<3x3xf32>) {

tensor<?x8xi1>) -> (tensor<8x10xf32>, tensor<?x8x10xf32>, tensor<8x10xf32>, tensor<8x10xf32>, tensor<f32>) attributes {tf._input_shapes = ["tfshape$dim { size: -1 } dim { size: 8 } dim { size: 8 }", "tfshape$dim { size: 8 } dim { size: 10 }", "tfshape$dim { size: 8 } dim { size: 10 }", "tfshape$unknown_rank: true", "tfshape$unknown_rank: false", "tfshape$unknown_rank: false", "tfshape$dim { size: -1 } dim { size: 8 }"], tf.api_implements = "lstm_b4e9f0e7-ac55-42bc-8ef2-8496419a608c", tf.api_preferred_device =...

  %0 = "tf.ReadVariableOp"(%handle) {_output_shapes = ["tfshape$dim { size: 3 }"], device = "/device:CPU:0", dtype = f32} : (tensor<!tf_type.resource<tensor<3xf32>>>) -> tensor<3xf32>
  %1 = "tf.ReadVariableOp"(%arg1) {_output_shapes = ["tfshape$dim { size: 1 } dim { size: 3 }"], device = "/device:CPU:0", dtype = f32} : (tensor<!tf_type.resource<tensor<1x3xf32>>>) -> tensor<1x3xf32>

      %outputs_3, %control_4 = tf_executor.island wraps "tf.Placeholder"() {device = "", dtype = "tfdtype$DT_FLOAT", name = "x", shape = "tfshape$dim { }"} : () -> tensor<0xf32>
      %outputs_5, %control_6 = tf_executor.island wraps "tf.TPUReplicatedInput"(%outputs_3) {N = 1 : i64, T = "tfdtype$DT_FLOAT", device = "", name = "input0"} : (tensor<0xf32>) -> tensor<0xf32>

      -> (tensor<3x3xf32> {tf_saved_model.index_path = []}) {
  // CHECK: {{%.*}} = corert.get_op_handler %arg0 "/device:GPU:0"
  %2 = "tf.MatMul"(%arg0, %arg1) {T = f32, _output_shapes = ["tfshape$dim { size: 3 } dim { size: 3 }"], device = "/device:GPU:0", transpose_a = false, transpose_b = false} : (tensor<3x1xf32>, tensor<1x3xf32>) -> tensor<3x3xf32>
  func.return %2 : tensor<3x3xf32>

  // CHECK: %[[RES:.*]] = "tf.Const"()
  // CHECK: return %[[RES]]
  %0 = "tf.Const"() {value = dense<4.200000e+01> : tensor<f32>} : () -> tensor<f32>
  %1 = "tf.VarHandleOp"() {container = "", shape = "tfshape$", shared_name = "x"} : () -> tensor<!tf_type.resource<tensor<f32>>>
  "tf.AssignVariableOp"(%1, %0) : (tensor<!tf_type.resource<tensor<f32>>>, tensor<f32>) -> ()
  func.return
}

// -----

namespace mangling_util {
namespace {

using ::mlir::tfg::mangling_util::PrintShortTextProto;

const char kAttributePrefix[] = "tf.";
const char kDataTypePrefix[] = "tfdtype$";
const char kTensorShapePrefix[] = "tfshape$";
const char kTensorPrefix[] = "tftensor$";

}  // namespace

string MangleAttributeName(absl::string_view str) {
  return absl::StrCat(kAttributePrefix, str);
}

    %outputs_0, %control_0 = tf_executor.island wraps "tf.Placeholder"() {device = "", dtype = "tfdtype$DT_FLOAT", name = "y", shape = "tfshape$dim { }"} : () -> tensor<0xf32>
    %outputs_1, %control_1 = tf_executor.island wraps "tf.TPUReplicatedInput"(%outputs_0) {N = 1 : i64, T = "tfdtype$DT_FLOAT", device = "", name = "input1"} : (tensor<0xf32>) -> tensor<0xf32>

  %cst_0 = arith.constant dense<1> : tensor<3xi32>
  %0 = "tf.Squeeze"(%arg0) {T = f32, _output_shapes = ["tfshape$dim { size: 4 } dim { size: 64 } dim { size: 64 }"], device = "", squeeze_dims = []} : (tensor<4x64x64x1xf32>) -> tensor<4x64x64xf32>

  // CHECK-DAG: [[LINSPACE:%.*]] = chlo.broadcast_add [[MUL]], [[START]] {broadcast_dimensions = array<i64>}
  // CHECK: return [[LINSPACE]]
  %0 = "tf.Const"() {_output_shapes = ["tfshape$"], device = "", dtype = i32, value = dense<4> : tensor<i32>} : () -> tensor<i32>
  %1 = "tf.LinSpace"(%arg0, %arg1, %0) : (tensor<f32>, tensor<f32>, tensor<i32>) -> tensor<4xf32>
  func.return %1 : tensor<4xf32>
}