#include "tensorflow/core/tfrt/utils/fallback_tensor.h"
#include "tfrt/bef/bef_buffer.h"  // from @tf_runtime
#include "tfrt/bef_converter/mlir_to_bef.h"  // from @tf_runtime
#include "tfrt/bef_executor/bef_file.h"  // from @tf_runtime
#include "tfrt/host_context/async_value.h"  // from @tf_runtime
#include "tfrt/host_context/chain.h"  // from @tf_runtime
#include "tfrt/host_context/execution_context.h"  // from @tf_runtime

// CHECK: TPUCompile
// CHECK: TPUExecute
// CHECK-NOT: func @_func
    tf_executor.graph {
      %outputs, %control = tf_executor.island wraps "arith.constant"() {value = dense<2.000000e+00> : tensor<f32>} : () -> tensor<f32>
      %outputs_0, %control_1 = tf_executor.island wraps "arith.constant"() {value = dense<3.000000e+00> : tensor<f32>} : () -> tensor<f32>

    %0 = tf_executor.graph {
      %1, %c0 = tf_executor.island wraps "tf.Const"() {value = dense<1> : tensor<i1>} : () -> tensor<i1>
      tf_executor.fetch %1 : tensor<i1>
    }
    return %0 : tensor<i1>
  }

  func.func private @while_body(%arg0: tensor<i32>, %arg1: tensor<i32>) -> (tensor<i32>, tensor<i32>) {
    tf_executor.graph {

  // CHECK-NOT: _replication_info =
  %0 = tf_executor.graph {
    %outputs, %control = tf_executor.island wraps "tf.GuaranteeConst"(%arg1) {T = f32, device = ""} : (tensor<f32>) -> tensor<f32>
    %outputs_0, %control_1 = tf_executor.island wraps "tf.GuaranteeConst"(%arg2) {T = f32, device = ""} : (tensor<f32>) -> tensor<f32>
    %control_2 = tf_executor.island wraps "tf.NoOp"() {_pivot_for_cluster = "cluster", device = ""} : () -> ()

    tf_executor.graph {
      %control = tf_executor.island {
        "tf.StatefulPartitionedCall"(%arg0) <{config = "", config_proto = "", executor_type = "", f = @partitioned}> : (tensor<*x!tf_type.resource>) -> ()
        tf_executor.yield
      }
      tf_executor.fetch %control : !tf_executor.control
    }
    return
  }

  // CHECK: tf_executor.graph
  tf_executor.graph {
    %handle0, %handle0_control = tf_executor.island wraps "tf.SomeOp"() : () -> !tf_res
    %handle1, %handle1_control = tf_executor.island wraps "tf.SomeOp"() : () -> !tf_res
    %control_A = tf_executor.island wraps "tf.OpA"() : () -> ()
    %while_out:2, %while_control = tf_executor.island(%control_A) wraps "tf.While"(

    %0 = tf_executor.graph {
      %outputs, %control = tf_executor.island wraps "tf.Identity"(%arg0) {device = ""} : (tensor<2x3xi32>) -> tensor<2x3xi32>
      %outputs_0, %control_1 = tf_executor.island wraps "tf.XlaSharding"(%outputs) {_XlaSharding = "", device = "", sharding = "", unspecified_dims = []} : (tensor<2x3xi32>) -> tensor<2x3xi32>

        %4 = "tf.B"(%2) : (tensor<?xi32>) -> tensor<?xi32>
        tf_device.return %4 : tensor<?xi32>
      }) {} : () -> tensor<?xi32>

      // CHECK: tf_executor.yield %[[CLUSTER_OUTPUT]]
      tf_executor.yield %3 : tensor<?xi32>
    }
    tf_executor.fetch %1#0 : tensor<?xi32>
  }
  func.return %0 : tensor<?xi32>
}

// CHECK: func private @[[CLUSTER]]

func.func @nested_entry_functions() attributes {tf.entry_function = {}} {
  tf_executor.graph {
     %control = tf_executor.island wraps "tf.StatefulPartitionedCall"() {config = "", config_proto = "", device = "/device:CPU:0", executor_type = "", f = @func} : () -> ()
     tf_executor.fetch
  }
  func.return
}

func.func @func() attributes {tf.entry_function = {}} {
  func.return
}

// tf_executor.island {_parallel_execution_ids=r0:0,p0:0} wraps "tf.OpC"
// tf_executor.island {_parallel_execution_ids=r0:0,p0:1} wraps "tf.OpD"
// tf_executor.island {_parallel_execution_ids=r0:1} wraps "tf.OpB"
// tf_executor.island {_parallel_execution_ids=r0:1,p0:0} wraps "tf.OpC"
// tf_executor.island {_parallel_execution_ids=r0:1,p0:1} wraps "tf.OpD"
//
// "tf.OpA" will not have `_parallel_execution_ids` attr,