%outputs_50 =  "tf.Mul"(%outputs_30, %outputs_48) {device = ""} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>
  %outputs_52 =  "tf.Reshape"(%outputs_50, %outputs_0) {device = ""} : (tensor<*xf32>, tensor<2xi32>) -> tensor<*xf32>
  %outputs_54 =  "tf.MatMul"(%outputs_40, %outputs_52) {device = "", transpose_a = false, transpose_b = false} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32>

  %cst_zero = arith.constant dense<0.0> : tensor<f32>
  %cst_one = arith.constant dense<1.0> : tensor<f32>

  // CHECK-DAG: %[[RES_NO_FOLD:.*]] = "tf.Pow"(%arg0, %arg1)
  %0 = "tf.Pow"(%arg0, %arg1) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>

  // CHECK-DAG: %[[POW_ZERO:.*]] = "tf.Const"() <{value = dense<1.000000e+00> : tensor<4xf32>}> : () -> tensor<4xf32>

// CHECK-LABEL: func @init
func.func @init(%arg0: tensor<4xf32>) -> tensor<7xf32> {
    %0 = builtin.unrealized_conversion_cast to tensor<7xf32>
    return %0 : tensor<7xf32>
}

// CHECK-LABEL: func @next
func.func @next(%arg0: tensor<7xf32>, %arg1: tensor<3xf32>) -> tensor<6xf32> {
    %0 = builtin.unrealized_conversion_cast to tensor<6xf32>
    return %0 : tensor<6xf32>
}

// CHECK-LABEL: func @finalize

  %1 = "tf.Mul"(%arg1, %arg1) : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xf32>
  %2 = "tf.Div"(%arg1, %arg1) : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xf32>

  // CHECK: "tf.If"{{.+}}else_branch = @tf.IfRegion_else{{.+}}then_branch = @tf.IfRegion_then
  %3 = "tf.IfRegion"(%arg0) ({
     %4 = "tf.Select"(%arg0, %0, %1) : (tensor<i1>, tensor<2xf32>, tensor<2xf32>) -> tensor<2xf32>
     "tf.Yield"(%4) : (tensor<2xf32>) -> ()

  // expected-remark@above {{Sinks: {}}}
}

// -----

func.func @add(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  // expected-remark@above {{ID: 2}}
  %sum = "tf.Add"(%arg0, %arg1) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  // expected-remark@above {{ID: 0}}
  func.return %sum : tensor<1xf32>
  // expected-remark@above {{ID: 1}}
  // expected-remark@above {{Sinks: {}}}
}

  # CHECK-SAME:   %arg0: tensor<1xf32> {tf._user_specified_name = "d", tf_saved_model.index_path = [0, "a"]},
  # CHECK-SAME:   %arg1: tensor<2xf32> {tf._user_specified_name = "d", tf_saved_model.index_path = [0, "b"]},
  # CHECK-SAME:   %arg2: tensor<3xf32> {tf._user_specified_name = "d", tf_saved_model.index_path = [0, "c"]},
  # CHECK-SAME:   %arg3: tensor<4xf32> {tf._user_specified_name = "d", tf_saved_model.index_path = [0, "x"]},

}

func.func @FusingaddRelu(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  %0 = "tf.Add"(%arg0, %arg1) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %1 = "tf.Add"(%arg0, %0) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  %2 = "tf.Relu"(%1) : (tensor<1xf32>) -> tensor<1xf32>
  %3 = "tf.Relu"(%arg0) : (tensor<1xf32>) -> tensor<1xf32>
  %4 = "tf.Add"(%3, %2) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>

  %10 = "tfl.pseudo_const"() {value = dense<0.000000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
  %11 = "tfl.pseudo_const"() {value = dense<1.000000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
  %recurrent_input = "tfl.pseudo_const"() {value = dense<0.000000e+00> : tensor<1x3xf32>} : () -> tensor<1x3xf32>
  %recurrent_stats = "quantfork.stats"(%recurrent_input) {layerStats = dense<[0.0, 1.0]> : tensor<2xf32>} : (tensor<1x3xf32>) -> tensor<1x3xf32>

// CHECK:           return %[[VAL_2]], %[[VAL_3]] : tensor<4xf32>, tensor<4xf32>
// CHECK:         }
func.func @broadcast_atan2(%arg0: tensor<1xf32>, %arg1: tensor<4xf32>) -> (tensor<4xf32>, tensor<4xf32>) {
  %0 = "mhlo.broadcast_in_dim"(%arg0) <{broadcast_dimensions = dense<[0]> : tensor<1xi64>}> : (tensor<1xf32>) -> tensor<4xf32>
  %1 = mhlo.atan2 %0, %arg1 : tensor<4xf32>
  %2 = mhlo.atan2 %arg1, %0 : tensor<4xf32>

  func.return %0 : tensor<*xf32>
}

// CHECK-NOT: "tfl.reshape"
// CHECK: "tfl.pack"

// -----

func.func @squaredDifference(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>) -> tensor<4xf32> {
  %0 = "tfl.squared_difference"(%arg0, %arg1) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
  func.return %0 : tensor<4xf32>
}