"tf.AssignVariableOp"(%arg0, %device_output) : (!tf_res_i32, tensor<i32>) -> ()
  "tf.SomeOp"(%device_output) : (tensor<i32>) -> ()
  func.return
}

// CHECK-LABEL: func @device_func_1
// CHECK-NOT: tf.aliasing_output
func.func @device_func_1(%arg0: tensor<i32>) -> tensor<i32> {
  func.return %arg0 : tensor<i32>
}

      %a = "tf.opA"(%arg0) : (tensor<i32>) -> tensor<i32>
      %b = "tf.opB"(%a) : (tensor<i32>) -> tensor<i32>
      tf_executor.yield %a, %b : tensor<i32>, tensor<i32>
    }
    %output1_0, %output1_1, %control1 = tf_executor.island {
      %c = "tf.opC"(%output0_0) : (tensor<i32>) -> tensor<i32>
      %d = "tf.opD"(%c) : (tensor<i32>) -> tensor<i32>
      tf_executor.yield %c, %d : tensor<i32>, tensor<i32>
    }

    %outputs_7, %control_7 = tf_executor.island wraps "tf.AddV2"(%outputs_5, %outputs) {_xla_compile_device_type = "TPU", _replication_info = "cluster1"} : (tensor<i32>, tensor<i32>) -> tensor<i32>
    %outputs_8, %control_8 = tf_executor.island wraps "tf.AddV2"(%outputs_0, %outputs_3) {_xla_compile_device_type = "TPU"} : (tensor<i32>, tensor<i32>) -> tensor<i32>
    tf_executor.fetch
  }

    func.return %0 : tensor<i32>
  }
  func.func @branch_1(%arg0: tensor<i32>) -> tensor<i32> {
    %0 = "tf.some_op"(%arg0) {_xla_compile_device_type = "TPU", _replication_info = "cluster"} : (tensor<i32>) -> tensor<i32>
    func.return %0 : tensor<i32>
  }
  func.func @branch_2(%arg0: tensor<i32>) -> tensor<i32> {

// CHECK-DAG:   [[ITEM_RANK:%.*]] = "tf.Rank"([[ITEM]]) : (tensor<10xf32>) -> tensor<i32>
// CHECK-DAG:   [[ZERO:%cst.*]] = arith.constant dense<0> : tensor<i32>
// CHECK-DAG:   [[VECTOR_RANK:%.*]] = "tf.ExpandDims"([[ITEM_RANK]], [[ZERO]]) : (tensor<i32>, tensor<i32>) -> tensor<1xi32>

func.func private @then(%x: tensor<i32>, %y: tensor<i32>) -> (tensor<i32>, tensor<i32>) {
  func.return %x, %x : tensor<i32>, tensor<i32>
}

func.func private @else(%x: tensor<i32>, %y: tensor<i32>) -> (tensor<i32>, tensor<i32>) {
  func.return %y, %y : tensor<i32>, tensor<i32>
}

// CHECK-LABEL: then/tfrt_dedup_results
// CHECK: return {{%.*}} : tensor<i32>

// CHECK-LABEL: else/tfrt_dedup_results

func.func @island_no_wrap_print_multiple_ops(%arg0: tensor<i32>, %arg1: tensor<f32>) {
  tf_executor.graph {
    // CHECK: tf_executor.island
    // CHECK-NOT: wraps
    %0:3 = tf_executor.island {
      %1:2 = "tf.IdentityN"(%arg0, %arg1) : (tensor<i32>, tensor<f32>) -> (tensor<i32>, tensor<f32>) loc("identity@some_function")
      %2:2 = "tf.IdentityN"(%1#0, %1#1) : (tensor<i32>, tensor<f32>) -> (tensor<i32>, tensor<f32>) loc("identity@some_function")

  %zps2 = "tf.Const"() { value = dense<2> : tensor<i32> } : () -> tensor<i32>
  %zps4 = "tf.Const"() { value = dense<4> : tensor<i32> } : () -> tensor<i32>
  %one = "tf.Const"() { value = dense<1> : tensor<i32> } : () -> tensor<i32>

  // CHECK: %[[QUANT0:.*]] = mhlo.uniform_quantize %[[ARG:.*]] : (tensor<2x2xf32>) -> tensor<2x2x!quant.uniform<i8:f32, 1.000000e+00:2>>

  %one = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
  %elem_shape = "tf.Const"() {value = dense<[-1, 1]> : tensor<2xi32>} : () -> tensor<2xi32>
  %size = "tf.Const"() {value = dense<10> : tensor<i32>} : () -> tensor<i32>
  // CHECK: TensorListReserve
  // CHECK: TensorListReserve
  %tl = "tf.TensorListReserve"(%elem_shape, %size) : (tensor<2xi32>, tensor<i32>) -> tensor<!tf_type.variant<tensor<?x1xf32>>>

  // CHECK: "tfl.average_pool_2d"(%{{.*}}) <{filter_height = 3 : i32, filter_width = 6 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 3 : i32, stride_w = 1 : i32}> : (tensor<1x6x6x16xf32>) -> tensor<1x1x1x16xf32>
  %0 = "tfl.average_pool_2d"(%arg0) {filter_height = 3 : i32, filter_width = 6 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 3 : i32, stride_w = 1 : i32} : (tensor<1x6x6x16xf32>) -> tensor<1x1x1x16xf32> loc("avgpool")