%2 = "tf.AddN"(%arg0, %1, %1) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %3 = "tf.AddN"(%1, %arg0, %1) : (tensor<2xi32>, tensor<2xi32> , tensor<2xi32>) -> tensor<2xi32>
  %4 = "tf.AddN"(%1, %1) : (tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %5 = "tf.AddN"(%arg0, %1, %0) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  func.return %2, %3, %4, %5: tensor<2xi32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>
}

// version.

// CHECK-LABEL: allows_chlo
func.func @allows_chlo(%arg0: tensor<1x32x10x32xi32>, %arg1: tensor<32xi32>) -> tensor<1x32x10x32xi32> {
  %0 = "chlo.broadcast_add"(%arg0, %arg1) {broadcast_dimensions = array<i64: 3>} : (tensor<1x32x10x32xi32>, tensor<32xi32>) -> tensor<1x32x10x32xi32>
  func.return %0 : tensor<1x32x10x32xi32>

  %outputs_0 = "tf.Const"() {value = dense<224> : tensor<2xi32>} : () -> tensor<2xi32>
  %outputs_2 = "tf.Const"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
  %outputs_4 = "tf.Const"() {value = dense<1> : tensor<1xi32>} : () -> tensor<1xi32>
  %outputs_6 = "tf.Const"() {value = dense<0> : tensor<1xi32>} : () -> tensor<1xi32>

  "tf_saved_model.asset"() {sym_name = "z", filename = "file"} : () -> ()
  func.func @serving_default(
      %arg0: tensor<1x3xi32> {tf_saved_model.index_path = ["x"]},
      %arg1: tensor<!tf_type.resource<tensor<3x1xi32>>> {tf_saved_model.bound_input = @y},
      %arg2: tensor<!tf_type.string> {tf_saved_model.bound_input = @z}

  // CHECK-DAG: %[[X_SHAPE:.+]] = shape.shape_of %[[X]] : tensor<3x4x256x6xf32> -> tensor<4xindex>
  // CHECK-DAG: %[[EPS:.+]] = mhlo.constant dense<1.000000e+00> : tensor<256xf32>
  // CHECK-DAG: %[[MEAN:.+]] = "tf.Mean"(%arg0, %[[CST_AXIS]]) <{keep_dims = false}> : (tensor<3x4x256x6xf32>, tensor<3xi32>) -> tensor<256xf32>

    %1 = stablehlo.constant dense<1> : tensor<1x1xi8>  // Input zero point.
    %2 = stablehlo.constant dense<5> : tensor<2x3xi32>  // Quantized filter - the pattern expects i8 but i32 is given.
    %3 = stablehlo.constant dense<4> : tensor<1x3xi32>  // Precalculated z1 * q2.
    %4 = stablehlo.constant dense<3.000000e+03> : tensor<1x3xf32>  // Merged scale: s1 * s2.

  func.return %0 : tensor<2xi32>
}

// -----

// CHECK-LABEL: listFromTensor
func.func @listFromTensor(%tensor: tensor<3xi32>, %shape : tensor<?xi32>) -> tensor<!tf_type.variant<tensor<i32>>> {
  %0 = "tf.TensorListFromTensor"(%tensor, %shape) : (tensor<3xi32>, tensor<?xi32>) -> tensor<!tf_type.variant<tensor<i32>>>

func.func @main(
    %arg0: tensor<10x10xf32>, %scale: tensor<10xf32>, %zp: tensor<10xi32>
  ) -> tensor<10x10xi8> {
  %0 = "tf.UniformQuantize"(%arg0, %scale, %zp) {
    quantization_axis = 1 : i64,
    quantization_min_val = -128 : i64,
    quantization_max_val = 127 : i64
  } : (tensor<10x10xf32>, tensor<10xf32>, tensor<10xi32>) -> tensor<10x10x!tf_type.qint8>

func.func @testNotTfDialect(%arg0: tensor<1x32x10x32xi32>, %arg1: tensor<32xi32>) -> tensor<1x32x10x32xi32> {
 // expected-error@below {{op is in dialect chlo not in tf functional dialect}}
  %0 = "chlo.broadcast_add"(%arg0, %arg1) {broadcast_dimensions = array<i64: 3>} : (tensor<1x32x10x32xi32>, tensor<32xi32>) -> tensor<1x32x10x32xi32>
  func.return %0 : tensor<1x32x10x32xi32>
}

// -----

func.func private @simple_callee(%arg0: tensor<2xi32>) -> tensor<2xi32> {
  %graph = tf_executor.graph {
    tf_executor.fetch %arg0 : tensor<2xi32>
  }
  func.return %graph#0 : tensor<2xi32>
}

// Test with a TPUPartitionedCallOp that is not inlined.
func.func @dont_inline_tpu_partitioned_call(%arg0: tensor<2xi32> {tf._user_specified_name = "inputs_0"}, %arg1: tensor<2xi32> {tf._user_specified_name = "inputs_1"}) -> tensor<2xi32> {