// Just to introduce an extra use for %cst.
  %1 = "tf.AddV2"(%cst, %cst) {device = ""} : (tensor<1xi32>, tensor<1xi32>) -> tensor<1xi32>

  return %0 : tensor<16xf32>
}
// Check that the constant is cloned with same value.
// CHECK-DAG: %[[CST_0:.*]] = "tf.Const"()
// CHECK-SAME: value = dense<1> : tensor<1xi32>

}

func.func @div(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  %0 = "tf.Div"(%arg0, %arg1) : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>
  func.return %0: tensor<1xf32>

// CHECK-LABEL: div
// CHECK:  tfl.div %arg0, %arg1 {fused_activation_function = "NONE"} : tensor<1xf32>
// CHECK:  return
}

func.func @squaredDifferenceRelu(tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32> {

  %0 = stablehlo.constant dense<2.000000e+00> : tensor<167x64xf32>
  %1 = stablehlo.constant dense<2.000000e+00> : tensor<64xf32>
  %2 = stablehlo.dot_general %arg0, %0, contracting_dims = [2] x [0], precision = [DEFAULT, DEFAULT] : (tensor<1x1x167xf32>, tensor<167x64xf32>) -> tensor<1x1x64xf32>
  %3 = stablehlo.broadcast_in_dim %1, dims = [2] : (tensor<64xf32>) -> tensor<1x1x64xf32>
  %4 = stablehlo.add %2, %3 : tensor<1x1x64xf32>
  func.return %4: tensor<1x1x64xf32>
}

  // CHECK-NEXT: %[[ADDN:.*]] = "tf.AddN"(%[[UPDATE]], %[[BROADCAST]]) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10xf32>
  %addn = "tf.AddN"(%set, %tl) : (tensor<!tf_type.variant<tensor<f32>>>, tensor<!tf_type.variant<tensor<f32>>>) -> tensor<!tf_type.variant<tensor<f32>>>
  // CHECK-NEXT: %[[ZEROS_LIKE:.*]] = "tf.ZerosLike"(%[[ADDN]]) : (tensor<10xf32>) -> tensor<10xf32>

  func.return %0#0 : tensor<?x?x?x?xf32>
}

// -----

// CHECK-LABEL: fusedBatchNormV3_training_dynamic_unsupported1

        %3 = "tf.Mul"(%arg0, %0) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>
        %4 = "tf.Mul"(%3, %0) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>
        %5 = "tf.Mul"(%4, %1) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>
        %6 = "tf.Mul"(%5, %2) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>
        tf_device.return %6 : tensor<16xf32>
      }) {} : () -> tensor<16xf32>
      tf_executor.yield %3 : tensor<16xf32>
    }

  %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>

  %0 = "tfl.unidirectional_sequence_lstm"(%arg0,...

  %2 = "tf.Mul"(%1, %value) {T = "tfdtype$DT_FLOAT"} : (tensor<256x8x7x16xf32>, tensor<16xf32>) -> tensor<256x8x7x16xf32>
  func.return %2 : tensor<256x8x7x16xf32>

// CHECK-DAG: %[[cst:.*]] = "tf.Const{{.*}} dense<8.000000e+00> : tensor<3x3x3x16xf32>
// CHECK-DAG: %[[cst_0:.*]] = "tf.Const{{.*}} dense<1.200000e+01> : tensor<16xf32>

// CHECK: Placeholder
// CHECK: key: "type"
// CHECK: type: DT_INT8

func.func @main(%arg0 : tensor<16xf32>) {
  tf_executor.graph {
    %1:2 = tf_executor.island wraps "tf.MlirPassthroughOp"(%arg0) {extra_type_attr = [tensor<5xi32>, tensor<16xf32>], Tinputs = [tensor<16xf32>], Toutputs = [tensor<16xf32>], mlir_module = ""} : (tensor<16xf32>) -> tensor<16xf32>
    tf_executor.fetch
  }
  func.return
}

func.func @plain() {
  tf_executor.graph {