#include "xla/tsl/c/tsl_status_internal.h"
#include "xla/tsl/distributed_runtime/coordination/coordination_service_agent.h"
#include "tensorflow/core/common_runtime/composite_device.h"
#include "tensorflow/core/common_runtime/device.h"
#include "tensorflow/core/common_runtime/eager/eager_operation.h"
#include "tensorflow/core/distributed_runtime/coordination/coordination_service_error_util.h"
#include "tensorflow/core/framework/function.h"

                            const char* const* underlying_devices,
                            int num_underlying_devices,
                            TFE_CustomDevice* device, void** device_info) {
  device->copy_tensor_to_device = &CopyToParallelDevice;
  device->copy_tensor_from_device = &CopyTensorFromParallelDevice;
  device->delete_device = &DeleteParallelDevice;
  device->execute = &ParallelDeviceExecute;

  TFE_CustomDevice custom_device;
  custom_device.copy_tensor_to_device = &CopyToLoggingDevice;
  custom_device.copy_tensor_from_device = &CopyTensorFromLoggingDevice;
  custom_device.delete_device = &DeleteLoggingDevice;
  custom_device.execute = &LoggingDeviceExecute;
  LoggingDevice* device = new LoggingDevice;
  device->arrived_flag = arrived_flag;
  device->executed_flag = executed_flag;

  const char* device_name = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  const char* first_device_name =
      "/job:localhost/replica:0/task:0/device:CPU:0";
  const char* second_device_name =
      "/job:localhost/replica:0/task:0/device:CPU:1";
  std::array<const char*, 2> underlying_devices{first_device_name,
                                                second_device_name};

                                  absl::optional<int64_t> step_id) const {
  for (int device_index = 0; device_index < underlying_devices_.size();
       ++device_index) {
    DeviceThread* device_thread = device_threads_[device_index].get();
    std::vector<TFE_TensorHandle*> device_inputs;
    device_inputs.reserve(inputs.size());
    for (int input_index = 0; input_index < inputs.size(); ++input_index) {

// Create and modify a variable placed on a parallel device which composes
// `first_device` and `second_device`.
void BasicTestsForTwoDevices(TFE_Context* context, const char* first_device,
                             const char* second_device) {
  // Register the custom device
  std::unique_ptr<TF_Status, decltype(&TF_DeleteStatus)> status(
      TF_NewStatus(), TF_DeleteStatus);

  const string composite_device_name =
      "/job:localhost/replica:0/task:0/device:COMPOSITE:0";
  EXPECT_EQ(TFE_TensorHandleDeviceName(packed_handle, status),
            composite_device_name);
  EXPECT_EQ(TFE_TensorHandleBackingDeviceName(packed_handle, status),
            composite_device_name);
  ASSERT_EQ(TF_GetCode(status), TF_OK) << TF_Message(status);

    t.reset(x_raw);
  }

  AbstractOperationPtr check_numerics_op(ctx->CreateOperation());
  ForwardOperation forward_op;
  Status s = Reset(check_numerics_op.get(), "CheckNumerics",
                   /*raw_device_name=*/nullptr, &forward_op);
  ASSERT_EQ(errors::OK, s.code()) << s.message();
  if (isa<TracingOperation>(check_numerics_op.get())) {
    s = dyn_cast<TracingOperation>(check_numerics_op.get())

      TF_DeleteStatus(status);
      return "";
    }
    TF_DeleteStatus(status);

    // We add a random value into `test_dir` to ensures that two consecutive
    // runs are unlikely to clash.
    std::random_device rd;
    std::mt19937 gen(rd());
    std::uniform_int_distribution<> distribution;
    std::string rng_val = std::to_string(distribution(gen));
    return tensorflow::io::JoinPath(std::string(test_dir), rng_val);
  } else {

                              const char* device_name, void* device_info,
                              TF_Status* status) {
  // Fill in default values for optional functionality.
  if (device.pack == nullptr) {
    device.pack = &DefaultCustomDevicePack;
  }
  auto custom_device = std::make_unique<tensorflow::CustomDeviceAPI>(
      ctx, device, device_info, device_name);