func.func @testDiv(%arg0: tensor<10x10x10xf32>, %arg1: tensor<10x10x10xf32>) -> tensor<10x10x10xf32> {
  // CHECK: _arithmetic_count = 1000 : i64
  %0 = "tfl.div"(%arg0, %arg1) {fused_activation_function = "NONE"} : (tensor<10x10x10xf32>, tensor<10x10x10xf32>) -> tensor<10x10x10xf32>
  func.return %0 : tensor<10x10x10xf32>
}

func.func @testDivBroadcast(%arg0: tensor<10x10x10xf32>, %arg1: tensor<10xf32>) -> tensor<10x10x10xf32> {

  %0 = "tfl.add"(%arg0, %arg1) {fused_activation_function = "RELU", tac.device = "GPU"} : (tensor<10x10x10xf32>, tensor<f32>) -> tensor<10x10x10xf32>
  func.return %0 : tensor<10x10x10xf32>
}

// -----

// CHECK: tac.cost = 2.000000e+03

}

// -----

func.func @func_0_CPU(%arg0: tensor<10x10x10xf32>, %arg1: tensor<10xf32>) -> tensor<10x10x10xf32> attributes {tac.device = "CPU", tac.interface_name = "func_0"} {
  // CHECK: tac.cost = 1.000000e+03
  %0 = "tfl.add"(%arg0, %arg1) {fused_activation_function = "RELU", tac.device = "CPU"} : (tensor<10x10x10xf32>, tensor<10xf32>) -> tensor<10x10x10xf32>
  func.return %0 : tensor<10x10x10xf32>
}

// -----

// CHECK-SAME:     interior_padding = dense<0> : tensor<3xi64>
// CHECK-SAME: }> : (tensor<10x10x10xf32>, tensor<f32>) -> tensor<10x10x10xf32>
// CHECK:      return %[[RES]] : tensor<10x10x10xf32>
}

// -----

// CHECK-LABEL: testTwoConsecutivePadsTwoNegativeHighPad
func.func @testTwoConsecutivePadsTwoNegativeHighPad(%arg0: tensor<10x10x10xf32>) -> (tensor<9x9x9xf32>) {
  %0 = mhlo.constant dense<0.000000e+00> : tensor<f32>

  func.return %3 : tensor<10x10x10xf32>
  // CHECK:  return %arg0
}

func.func @NotFoldStridedSlice(%arg0 : tensor<10x10x10xf32>) -> tensor<9x9x9xf32> {
  %0 = "tfl.pseudo_const"() {value = dense<[9, 9, 9]> : tensor<3xi32>} : () -> tensor<3xi32>

  // CHECK: return %[[RES]]
}

// CHECK-LABEL: lower_rfft_to_rfft2d
func.func @lower_rfft_to_rfft2d(%input: tensor<10x20x30xf32>, %fft_len: tensor<1xi32>) -> tensor<10x20x30xcomplex<f64>> {
  %0 = "tf.RFFT"(%input, %fft_len) : (tensor<10x20x30xf32>, tensor<1xi32>) -> tensor<10x20x30xcomplex<f64>>
  func.return %0: tensor<10x20x30xcomplex<f64>>

func.func @testBatchMatMulV2InvalidBroadcastingBatchDimensionWithHigherXRank(%lhs: tensor<10x2x5x10xf32>, %rhs: tensor<10x10x10xf32>) {
  // expected-error @+1 {{found incompatible broadcast batch dimensions for lhs shape 'tensor<10x2x5x10xf32>' and rhs shape 'tensor<10x10x10xf32>'}}
  %0 = "tf.BatchMatMulV2"(%lhs, %rhs) : (tensor<10x2x5x10xf32>, tensor<10x10x10xf32>) -> tensor<10x10xf32>
}

// -----

  %5 = "tfl.log"(%4) : (tensor<10x1x10xf32>) -> tensor<10x1x10xf32>
  %6 = tfl.sub(%2, %5) {fused_activation_function = "NONE"} : (tensor<10x10x10xf32>, tensor<10x1x10xf32>) -> tensor<10x10x10xf32>
  return %6 : tensor<10x10x10xf32>
 // CHECK-NOT: "tfl.log_softmax"(%arg0) : (tensor<10x10x10f32>) -> tensor<10x10x10xf32>
}

  func.return %0 : tensor<1x10x10x32xi32>

// CHECK-LABEL: biasAddInt
// CHECK:  "tf.BiasAdd"(%arg0, %arg1)
}

func.func @squeezeAndReshape(%arg0: tensor<1x1x10xf32>, %arg1: tensor<?x10xf32>) -> i32 {
  %0 = "tf.Squeeze"(%arg0) {squeeze_dims = [0]} : (tensor<1x1x10xf32>) -> tensor<1x10xf32>
  %1 = "tf.Squeeze"(%arg1) : (tensor<?x10xf32>) -> tensor<*xf32>
  %2 = "tf.Const"() { value = dense<[2, 5]> : tensor<2xi32> } : () -> tensor<2xi32>

// CHECK-NEXT: return %[[add]] : tensor<256x8x7x16xf32>
}

// CHECK-LABEL: fold_cast_fft_to_rfft
func.func @fold_cast_fft_to_rfft(%arg0: tensor<10x20x30xf32>) -> tensor<10x20x30xcomplex<f32>> {
  %0 = "tf.Cast"(%arg0) : (tensor<10x20x30xf32>) -> tensor<10x20x30xcomplex<f32>>
  %1 = "tf.FFT"(%0) : (tensor<10x20x30xcomplex<f32>>) -> tensor<10x20x30xcomplex<f32>>
  func.return %1: tensor<10x20x30xcomplex<f32>>