func.func nested @_tpu_v1_compat_outlined_func1(%arg0: tensor<i1>, %arg1: tensor<f32>) -> (tensor<i1>, tensor<i32>) {
      %0 = "tf.opA"(%arg0) : (tensor<i1>) -> tensor<i1>
      %1 = "tf.opA"(%0) : (tensor<i1>) -> tensor<i1>
      %2 = "tf.SomeOp"(%arg0, %arg1) : (tensor<i1>, tensor<f32>) -> tensor<i32>
      func.return %1, %2 : tensor<i1>, tensor<i32>
    }
  }

    %dim:2 = tf_executor.island wraps "tf.Const"() {dtype = "tftype$DT_INT32", value = dense<0> : tensor<i32>} : () -> tensor<i32>
    %input:2 = tf_executor.island wraps "tf.Const"() {dtype = "tftype$DT_INT32", value = dense<1.0> : tensor<4x6xf32>} : () -> tensor<4x6xf32>
    %split:3 = tf_executor.island wraps "tf.Split"(%dim#0, %input#0) : (tensor<i32>, tensor<4x6xf32>) -> (tensor<2x6xf32>, tensor<2x6xf32>)
    tf_executor.fetch
  }
  func.return

    %1 = "tfl.conv_2d"(%arg0, %0, %arg2) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<1x224x224x3xf32>, tensor<32x3x3x3xf32>, tensor<32xf32>) -> tensor<1x112x112x32xf32>
    func.return %1 : tensor<1x112x112x32xf32>

// CHECK: %[[q0:.*]] = "tfl.quantize"(%arg2) <{qtype = tensor<32x!quant.uniform<i32:f32, 0.0078431372549019607>>}>

  %0 = mhlo.custom_call @custom_call.my_custom_op(%arg0) {
    api_version = 4 : i32,
    backend_config = {foo = "bar"}
  } : (tensor<1x4xf32>) -> (tensor<1x8xf32>)
  func.return %0 : tensor<1x8xf32>

  //       CHECK: %0 = mhlo.custom_call @custom_call.my_custom_op(%arg0) {
  //  CHECK-SAME:   api_version = 4 : i32,
  //  CHECK-SAME:   backend_config = {foo = "bar"}
  //  CHECK-SAME: } : (tensor<1x4xf32>) -> tensor<1x8xf32>

// CHECK: %[[CST:.*]] = "tfl.no_value"() <{value}> : () -> none
// CHECK: %[[RES0:.*]] = "tfl.conv_2d"(%arg0, %arg1, %[[CST]]) <{dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32}> : (tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, none) -> tensor<256x32x32x16xf32>
// CHECK: return %[[RES0]] : tensor<256x32x32x16xf32>

{
  "version": 3,

    %outputs, %control = tf_executor.island wraps "tf.Const"() {device = "", dtype = "tfdtype$DT_INT32", name = "Constant", value = dense<0> : tensor<i32>} : () -> tensor<i32>
    %outputs_0, %control_1 = tf_executor.island wraps "tf.LegacyCall"(%outputs) {_tpu_replicate = "cluster", device = "", f = @foo0} : (tensor<i32>) -> tensor<i32>
    tf_executor.fetch
  }
  func.return
}
func.func @foo0(%arg0: tensor<*xi32>) -> tensor<*xi32> {
  %0 = tf_executor.graph {

  %ch0 = tfrt.new.chain

  %ch1 = tfrt_fallback_async.createop(%ch0) key(0) device("/CPU:0") "tf.Const"()
         { dtype = i32, value = dense<[2]> : tensor<1xi32> } num_args(0)
  %ch2 = tfrt_fallback_async.createop(%ch1) key(1) device("/CPU:0")
         "tf.TestAsyncTfrtAsyncThread"() {T = i32} num_args(1)
  %ch3 = tfrt_fallback_async.createop(%ch2) key(2) device("/CPU:0")
         "tf.TestPrintThreadName"() num_args(0)

  %0 = "tfl.conv_2d"(%arg0, %arg1, %cst_0) {dilation_h_factor = 2 : i32, dilation_w_factor = 3 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 4 : i32, stride_w = 5 : i32} : (tensor<256x32x32x3xf16>, tensor<16x3x3x3xf16>, tensor<16xf16>) -> tensor<256x8x7x16xf16>

      %0 = "tfl.conv_2d"(%arg0, %arg1, %arg2) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU", padding = "VALID", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x4x4x3xf32>, tensor<3x1x1x3xf32>, tensor<3xf32>) -> tensor<1x4x4x3xf32>
      return %0 : tensor<1x4x4x3xf32>
    }

func.func @main(tensor<3x2xi32>) -> tensor<3x2xi32> {
^bb0(%arg0: tensor<3x2xi32>):
  %0 = "arith.constant"() {name = "Const2", value = dense<10> : tensor<i32>} : () -> tensor<i32>
  %1 = "tfl.add"(%0, %arg0) {fused_activation_function = "NONE", name = "add"} : (tensor<i32>, tensor<3x2xi32>) -> tensor<3x2xi32>
  func.return %1 : tensor<3x2xi32>