%0 = "tf.StatefulPartitionedCall"(%arg0) <{
      config = "", config_proto = "", executor_type = "", f = @some_func
    }> {
      _collective_manager_ids = [], device = ""
    } : (tensor<1x2x2x3xf32>) -> tensor<1x2x2x3xf32>
    // CHECK-SKIP: tf.StatefulPartitionedCall
    // CHECK-NOSKIP: call @some_func
    // CHECK-NOSKIP-NOT: tf.StatefulPartitionedCall
    %1 = "tf.PartitionedCall"(%0) <{

// CHECK: tf_device.cluster
// CHECK-NEXT: tf.StatefulPartitionedCall
// CHECK-NEXT: tf_device.return
// CHECK: _cluster_outlined_function_name = "stateful_pcall_func"
// CHECK-SAME: allow_soft_placement = true
// CHECK: tf.Const
// CHECK: tf.Add
func.func @xla_must_compile_true(%arg0: tensor<i32>) -> tensor<i32> attributes {tf.entry_function = {}} {

      // CHECK: %[[CST:.*]] = "tf.Const"
      %result = "tf.StatefulPartitionedCall"() {config = "", config_proto = "", executor_type = "", f = @simple_callee} : () -> tensor<2xi32>

      // Multi block regions may not.
      // CHECK-NEXT: %[[CALL:.*]] = "tf.StatefulPartitionedCall"

    key: "value"
    value {
      tensor {
        dtype: DT_FLOAT
        tensor_shape {
        }
        float_val: 0.0
      }
    }
  }
}
node {
  name: "statefulpartitionedcall"
  op: "StatefulPartitionedCall"
  input: "const"
  input: "args_1"
  attr {
    key: "Tin"
    value {
      list {
        type: DT_FLOAT
        type: DT_RESOURCE
      }
    }
  }

// CHECK-NEXT:      buffer: 6,
// CHECK-NEXT:      name: "StatefulPartitionedCall:0",
// CHECK-NEXT:      quantization: {
// CHECK-EMPTY:
// CHECK-NEXT:      },
// CHECK-NEXT:      shape_signature: [ -1, 5 ],
// CHECK-NEXT:      has_rank: true
// CHECK-NEXT:    }, {
// CHECK-NEXT:      shape: [ 1, 5 ],
// CHECK-NEXT:      buffer: 7,
// CHECK-NEXT:      name: "StatefulPartitionedCall:1",
// CHECK-NEXT:      quantization: {
// CHECK-EMPTY:

  op: "_Arg"
  attr {
    key: "T"
    value {
      type: DT_RESOURCE
    }
  }
  attr {
    key: "index"
    value {
      i: 0
    }
  }
}
node {
  name: "spc1"
  op: "StatefulPartitionedCall"
  input: "input0"
  input: "args_0"
  attr {
    key: "Tin"
    value {
      list {
        type: DT_FLOAT
        type: DT_RESOURCE
      }
    }
  }
  attr {
    key: "Tout"

func.func @convert_xla_must_compile(%arg0: tensor<i32>) -> tensor<i32> {
  // CHECK: "tf.StatefulPartitionedCall"(%arg0) <{config = "", config_proto = "", executor_type = "", f = @stateful_pcall_func}> {_xla_compile_device_type = "CPU", device = "/device:CPU:0"} : (tensor<i32>) -> tensor<i32>

  "tf.PartitionedCall"() {config = "",  config_proto = "", executor_type = "", f = @g} : () -> ()
  // CHECK: "tf.StatefulPartitionedCall"() <{config = "",  config_proto = "", executor_type = "", f = @[[NEWG:.+]]}> : () -> ()
  "tf.StatefulPartitionedCall"() {config = "",  config_proto = "", executor_type = "", f = @g} : () -> ()
  func.return
}

// CHECK: func.func @g()

    (tensor<1x10xf32> {tf_saved_model.index_path = ["output_0"]}, tensor<1x10xi64> {tf_saved_model.index_path = ["output_1"]}) attributes {tf.entry_function = {control_outputs = "", inputs = "serving_default_x:0", outputs = "StatefulPartitionedCall:0"}, tf_saved_model.exported_names = ["serving_default"]} {
    %handle_0 = "tf.VarHandleOp"() {container="c", shared_name="a"} : () -> tensor<!tf_type.resource<tensor<1x10xf32>>>

  # CHECK-SAME:   tensor<f32> {tf_saved_model.index_path = [1]})
  # CHECK-SAME: attributes{{.*}}tf_saved_model.exported_names = ["callee"]
  # CHECK:        "tf.StatefulPartitionedCall"{{.*}}f = @[[CALLEE_INTERNAL:[a-zA-Z_0-9]+]]
  #
  # CHECK:      func {{@[a-zA-Z_0-9]+}}(
  # CHECK-SAME:   %arg0: tensor<f32> {tf._user_specified_name = "x", tf_saved_model.index_path = [0]},