underlying_devices_.size());
  for (int component_index = 0; component_index < underlying_devices_.size();
       ++component_index) {
    if (!DeviceNameUtils::ParseFullName(underlying_devices_[component_index],
                                        &parsed_components[component_index]) ||
        !DeviceNameUtils::IsSameAddressSpace(
            underlying_devices_[component_index], underlying_devices_[0])) {

  // The number of devices operations run on.
  size_t num_underlying_devices() const { return underlying_devices_.size(); }

  // The devices operations run on.
  const std::vector<std::string>& underlying_devices() const {
    return underlying_devices_;
  }

  // Takes a description of a single operation being executed on the

  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};
  RegisterParallelDevice(context.get(), device_name, underlying_devices,
                         status.get());
  ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());

namespace parallel_device {

// Allocate a parallel device named `device_name` which forwards operations to
// `underlying_devices`, maintaining "parallel tensors" with components placed
// on each underlying device.
//
// For example if `device_name` is
//   "/job:localhost/replica:0/task:0/device:CUSTOM:0"
// and `underlying_devices` is
//   {"/job:localhost/replica:0/task:0/device:GPU:0",
//    "/job:localhost/replica:0/task:0/device:GPU:1"}

#include "tensorflow/c/tf_status.h"
#include "tensorflow/core/lib/gtl/cleanup.h"
#include "tensorflow/core/platform/test.h"

namespace {

struct LoggingDevice {
  tensorflow::string device_name;
  tensorflow::string underlying_device;
  // Set to true whenever a TensorHandle is copied onto the device
  bool* arrived_flag;
  // Set to true whenever an operation is executed
  bool* executed_flag;

  const char* device_name = "/job:localhost/replica:0/task:0/device:CUSTOM:0";
  std::array<const char*, 2> underlying_devices{first_device, second_device};
  RegisterParallelDevice(context.get(), device_name, underlying_devices,
                         status.get());
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());

  delete reinterpret_cast<NamedParallelDevice*>(device_info);
}

}  // namespace

void AllocateParallelDevice(const char* device_name,
                            const char* const* underlying_devices,
                            int num_underlying_devices,
                            TFE_CustomDevice* device, void** device_info) {
  device->copy_tensor_to_device = &CopyToParallelDevice;