func.func @mirror_pad(%arg0: tensor<2x3xcomplex<f64>>) -> tensor<4x7xcomplex<f64>> {
  %0 = mhlo.constant dense<[[1, 1], [2, 2]]> : tensor<2x2xi32>
  // NO_FALLBACK: tf.MirrorPad
  // SUPPORTED_FALLBACK_DEVICE-NOT: tf.MirrorPad
  // UNSPECIFIED_FALLBACK_DEVICE: tf.MirrorPad
  // UNSUPPORTED_FALLBACK_DEVICE: tf.MirrorPad
  %1 = "tf.MirrorPad"(%arg0, %0) {mode = "SYMMETRIC"} : (tensor<2x3xcomplex<f64>>, tensor<2x2xi32>) -> tensor<4x7xcomplex<f64>>

  // CHECK-LABEL: mirror_pad
  func.func @mirror_pad(%arg0: tensor<2x3xcomplex<f64>>) -> tensor<4x7xcomplex<f64>> {
    %0 = mhlo.constant dense<[[1, 1], [2, 2]]> : tensor<2x2xi32>
    // CHECK-NOT: tf.MirrorPad
    %1 = "tf.MirrorPad"(%arg0, %0) {mode = "SYMMETRIC"} : (tensor<2x3xcomplex<f64>>, tensor<2x2xi32>) -> tensor<4x7xcomplex<f64>>
    func.return %1 : tensor<4x7xcomplex<f64>>
  }

  // CHECK-LABEL: bucketize

  func.return %0 : tensor<5x4xf32>
}

func.func @batchmatmulv2_adj_complex(%arg0: tensor<2x5xcomplex<f32>>, %arg1: tensor<4x2xcomplex<f32>>) -> tensor<5x4xcomplex<f32>> {
// CHECK-LABEL:   func @batchmatmulv2_adj_complex(
// CHECK-SAME:                                    [[LHS:%.*]]: tensor<2x5xcomplex<f32>>, [[RHS:%.*]]: tensor<4x2xcomplex<f32>>) -> tensor<5x4xcomplex<f32>> {
// CHECK:           [[LHSRE:%.*]] = mhlo.real [[LHS]]

func.func @sigmoid_grad_complex(%arg0: tensor<2xcomplex<f32>>, %arg1: tensor<2xcomplex<f32>>) -> tensor<2xcomplex<f32>> {
  // CHECK-DAG: [[MUL0:%.+]] =  mhlo.multiply %arg1, %arg0 : tensor<2xcomplex<f32>>
  // CHECK-DAG: [[ONE:%.+]] = mhlo.constant dense<(1.000000e+00,0.000000e+00)> : tensor<2xcomplex<f32>>
  // CHECK-DAG: [[SUB:%.+]] =  mhlo.subtract [[ONE]], %arg0 : tensor<2xcomplex<f32>>

  %noncon3 = "tf.Complex"(%con1, %arg2) : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xcomplex<f32>>

  %res1 = "tf.MulNoNan"(%arg0, %arg1) : (tensor<2xcomplex<f32>>, tensor<2xcomplex<f32>>) -> tensor<2xcomplex<f32>>
  %res2 = "tf.DivNoNan"(%arg0, %noncon2) : (tensor<2xcomplex<f32>>, tensor<2xcomplex<f32>>) -> tensor<2xcomplex<f32>>

func.func @decompose_adam_with_complex_inputs(%arg0: tensor<!tf_type.resource<tensor<2xcomplex<f32>>>>, %arg1: tensor<!tf_type.resource<tensor<2xcomplex<f32>>>>, %arg2: tensor<!tf_type.resource<tensor<2xcomplex<f32>>>>, %arg3: tensor<complex<f32>>, %arg4: tensor<complex<f32>>, %arg5: tensor<complex<f32>>, %arg6: tensor<complex<f32>>, %arg7: tensor<complex<f32>>, %arg8: tensor<complex<f32>>, %arg9: tensor<2xcomplex<f32>>) attributes {tf.entry_function = {control_outputs = "Adam/update_Variable_1/ResourceApplyAdam",...

func.func @squared_difference_complex(%arg0: tensor<3xcomplex<f32>>, %arg1: tensor<3xcomplex<f32>>) -> tensor<3xcomplex<f32>> {
  // CHECK-DAG: [[R1:%.+]] = "tf.Sub"(%arg0, %arg1)
  // CHECK-DAG: [[R2:%.+]] = "tf.Conj"([[R1]])
  // CHECK-DAG: "tf.Mul"([[R1]], [[R2]])
  %1 = "tf.SquaredDifference"(%arg0, %arg1) : (tensor<3xcomplex<f32>>, tensor<3xcomplex<f32>>) -> tensor<3xcomplex<f32>>
  func.return %1 : tensor<3xcomplex<f32>>
}

// CHECK-SAME:                      %[[VAL_0:.*]]: tensor<2xcomplex<f32>>) -> tensor<2xf32> {
// CHECK:           %[[VAL_1:.*]] = "tf.ComplexAbs"(%[[VAL_0]]) : (tensor<2xcomplex<f32>>) -> tensor<2xf32>
// CHECK:           return %[[VAL_1]] : tensor<2xf32>
// CHECK:         }
func.func @complex_abs(%arg0: tensor<2xcomplex<f32>>) -> tensor<2xf32> {
  %0 = "mhlo.abs"(%arg0) : (tensor<2xcomplex<f32>>) -> tensor<2xf32>
  func.return %0 : tensor<2xf32>

func.func @testMinComplex(%arg0: tensor<4x8xcomplex<f32>>) -> tensor<4x1xcomplex<f32>> {
  %dimension = "tf.Const"() { value = dense<1> : tensor<1xi64> } : () -> tensor<1xi64>
  // expected-error@below {{'tf.Min' op operand #0 must be tensor of}}
  %0 = "tf.Min"(%arg0, %dimension) { keep_dims = true }: (tensor<4x8xcomplex<f32>>, tensor<1xi64>) -> tensor<4x1xcomplex<f32>>
  func.return %0 : tensor<4x1xcomplex<f32>>
}

// -----