// Two batch_size=2 batches get concatenated.
  %result_1 = tfrt_fallback_async.batch_function device("/device:CPU:0") @matmul_cpu (%a1, %b) {
      num_batch_threads = 1,
      max_batch_size = 4,
      allowed_batch_sizes = [4],
      batch_timeout_micros = 1000000,
      container = "container",
      shared_name = "shared_name",
      batching_queue = "batching_queue",

  // CHECK-SAME: batch_timeout_micros = 100000 : i64
  // CHECK-SAME: batching_queue = ""
  // CHECK-SAME: container = ""
  // CHECK-SAME: enable_large_batch_splitting = false
  // CHECK-SAME: max_batch_size = 6 : i64
  // CHECK-SAME: max_enqueued_batches = 10 : i64
  // CHECK-SAME: num_batch_threads = 2 : i64
  // CHECK-SAME: shared_name = "batch/"

    %1 = "tf.BatchFunction"(%0) {
        f = @called_by_batch_func,
        operandSegmentSizes = array<i32: 1, 0>,
        batch_timeout_micros = 1000,
        max_batch_size = 8,
        num_batch_threads = 2
      } : (tensor<!tf_type.resource<tensor<2xi32>>>) -> tensor<*xi32>
    return %1 : tensor<*xi32>
  }

  func.return %0#0 :...

  func.return %0#0 : tensor<*xi32>...

    value: {
      i: 100000
    }
  }
  attr: {
    key: "f"
    value: {
      func: {
        name: "inference_function"
      }
    }
  }
  attr: {
    key: "max_batch_size"
    value: {
      i: 32
    }
  }
  attr: {
    key: "num_batch_threads"
    value: {
      i: 16
    }
  }
  attr: {
    key: "shared_name"
    value: {
      s: ""

  %1 = "tf.BatchFunction"(%arg0, %0) {allowed_batch_sizes = [6], batch_timeout_micros = 100000 : i64, batching_queue = "", container = "", device = "/device:CPU:0", enable_large_batch_splitting = false, f = @batched_function, max_batch_size = 6 : i64, max_enqueued_batches = 10 : i64, num_batch_threads = 1 : i64, operandSegmentSizes = array<i32: 1, 1>, shared_name = "batch/"} : (tensor<1x3xf32>, tensor<!tf_type.resource<tensor<1x3xf32>>>) -> tensor<*xf32>

  // CHECK-SAME: batch_timeout_micros = 100000 : i64
  // CHECK-SAME: batching_queue = ""
  // CHECK-SAME: container = ""
  // CHECK-SAME: enable_large_batch_splitting = false
  // CHECK-SAME: max_batch_size = 6 : i64
  // CHECK-SAME: max_enqueued_batches = 10 : i64
  // CHECK-SAME: num_batch_threads = 1 : i64
  // CHECK-SAME: shared_name = "batch/"

      // CHECK-SAME: Tout = [f32]
      %outputs_2, %control_2 = tf_executor.island wraps "tf.BatchFunction"(%outputs_1) {batch_timeout_micros = 5000 : i64, batching_queue = "", container = "", f = @batched_func, max_batch_size = 256 : i64, num_batch_threads = 2 : i64, operandSegmentSizes = array<i32: 0, 1>, shared_name = ""} : (tensor<*x!tf_type.resource>) -> tensor<*xf32>
      tf_executor.fetch %outputs_2 : tensor<*xf32>
    }

// Tests that `tf.BatchFunction` is inlined.

func.func @serving_default(%arg0: tensor<1xf32>, %arg1: tensor<1xf32>) -> tensor<1xf32> {
  %0 = "tf.BatchFunction"(%arg0, %arg1) {f = @batched_func, num_batch_threads = 1 : i64, max_batch_size = 2 : i64, batch_timeout_micros = 10000 : i64, operandSegmentSizes = array<i32: 1, 1>} : (tensor<1xf32>, tensor<1xf32>) -> (tensor<1xf32>)
  return %0 : tensor<1xf32>
}