pytorch/torch/csrc/distributed/rpc/rref_impl.cpp

#include <torch/csrc/distributed/rpc/rref_impl.h>

#include <torch/csrc/distributed/autograd/rpc_messages/rpc_with_autograd.h>
#include <torch/csrc/distributed/autograd/utils.h>
#include <torch/csrc/distributed/rpc/python_rpc_handler.h>
#include <torch/csrc/distributed/rpc/rref_context.h>
#include <torch/csrc/distributed/rpc/rref_proto.h>
#include <torch/csrc/distributed/rpc/utils.h>

namespace torch {
namespace distributed {
namespace rpc {

namespace {

constexpr int OWNER_IDX = 0; // index of ownerId in the tuple
constexpr int RREFID_ON_IDX = 1; // index of RRefId.createdOn_ in the tuple
constexpr int RREFID_ID_IDX = 2; // index of RRefId.localId_ in the tuple
constexpr int FORKID_ON_IDX = 3; // index of ForkId.createdOn_ in the tuple
constexpr int FORKID_ID_IDX = 4; // index of ForkId.localId_ in the tuple
constexpr int PARENT_IDX = 5; // index of parent in the tuple

// NB: if more fields are added, make sure this field is also bumped
constexpr int RFD_TUPLE_SIZE = 6; // number of RRefForkData fields in py::tuple

template <typename T>
T& unwrapAutogradMessage(
    const Message& message,
    std::unique_ptr<RpcCommandBase>& response) {
  if (message.type() == MessageType::FORWARD_AUTOGRAD_RESP) {
    auto& rpcWithAutograd = static_cast<autograd::RpcWithAutograd&>(*response);

    // Attach 'recv' autograd function.
    addRecvRpcBackward(
        rpcWithAutograd.autogradMetadata(),
        rpcWithAutograd.tensors(),
        rpcWithAutograd.fromWorkerId());

    auto& wrappedRpc = rpcWithAutograd.wrappedRpc();
    return static_cast<T&>(wrappedRpc);
  } else {
    return static_cast<T&>(*response);
  }
}

} // namespace

std::atomic<local_id_t> RRefContext::nextLocalId_{0};

//////////////////////////  RRefForkData  /////////////////////////////////

RRefForkData::RRefForkData(
    worker_id_t ownerId,
    const RRefId& rrefId,
    const ForkId& forkId,
    worker_id_t parent)
    : ownerId_(ownerId), rrefId_(rrefId), forkId_(forkId), parent_(parent) {}

py::tuple RRefForkData::toPyTuple() const {
  return py::make_tuple(
      ownerId_,
      rrefId_.createdOn_,
      rrefId_.localId_,
      forkId_.createdOn_,
      forkId_.localId_,
      parent_);
}

RRefForkData RRefForkData::fromPyTuple(const py::tuple& t) {
  TORCH_INTERNAL_ASSERT(
      t.size() == RFD_TUPLE_SIZE,
      "Pickled RRefForkData must contain 6 numbers.");
  worker_id_t ownerId = t[OWNER_IDX].cast<worker_id_t>();
  // const reference will extend the lifetime of the temporary variable
  const RRefId& rrefId = RRefId(
      t[RREFID_ON_IDX].cast<worker_id_t>(),
      t[RREFID_ID_IDX].cast<local_id_t>());
  const RRefId& forkId = RRefId(
      t[FORKID_ON_IDX].cast<worker_id_t>(),
      t[FORKID_ID_IDX].cast<local_id_t>());
  worker_id_t parent = t[PARENT_IDX].cast<worker_id_t>();
  return RRefForkData(ownerId, rrefId, forkId, parent);
}

//////////////////////////////  RRef  /////////////////////////////////////

RRef::RRef(worker_id_t ownerId, const RRefId& rrefId)
    : RRefInterface(), ownerId_(ownerId), rrefId_(rrefId) {}

RRefForkData RRef::fork() const {
  auto& ctx = RRefContext::getInstance();
  return RRefForkData(
      ownerId_, rrefId_, ctx.genGloballyUniqueId(), ctx.getWorkerId());
}

//////////////////////////  UserRRef  /////////////////////////////////////

template <typename T>
UserRRef<T>::UserRRef(
    worker_id_t ownerId,
    const RRefId& rrefId,
    const ForkId& forkId)
    : RRef(ownerId, rrefId), forkId_(forkId) {
  // Do nothing,
  // (1) If this UserRRef is a fork of an existing RRef, RRefContext will send
  //     a RREF_FORK_REQUEST message to the owner.
  // (2) If this the creator UserRRef, ScriptRemoteCall or PythonRemoteCall will
  //     properly notify the owner.
}

template <typename T>
UserRRef<T>::~UserRRef() {
  try {
    RRefContext::getInstance().delUser(ownerId_, rrefId_, forkId_);
  } catch (const std::exception& ex) {
    LOG(ERROR) << "Error occurred when deleting UserRRef instance, "
               << "RRefId = " << rrefId_ << ", ForkId = " << forkId_ << " : "
               << ex.what();
  } catch (...) {
    LOG(ERROR) << "Error occurred when deleting UserRRef instance, "
               << "RRefId = " << rrefId_ << ", ForkId = " << forkId_ << " : "
               << "unknown error";
  }
}

template <typename T>
const ForkId& UserRRef<T>::forkId() const {
  return forkId_;
}

template <>
IValue UserRRef<IValue>::toHere() {
  auto agent = RpcAgent::getDefaultRpcAgent();

  // ScriptRRefFetchCall message always carries autograd context id even if
  // the message itself does not contain any tensor, because the response would
  // potentially contain tensors.
  auto futureResponse = autograd::sendMessageWithAutograd(
      *agent,
      agent->getWorkerInfo(ownerId_),
      ScriptRRefFetchCall(ownerId_, rrefId()).toMessage(),
      true /* forceGradRecording */);

  const Message& message = futureResponse->wait();
  auto response = deserializeResponse(message);
  auto& rfr = unwrapAutogradMessage<ScriptRRefFetchRet>(message, response);
  return rfr.values().front();
}

template <>
py::object UserRRef<py::object>::toHere() {
  auto agent = RpcAgent::getDefaultRpcAgent();

  // PythonRRefFetchCall message always carries autograd context id even if
  // the message itself does not contain any tensor, because the response would
  // potentially contain tensors.
  auto futureResponse = autograd::sendMessageWithAutograd(
      *agent,
      agent->getWorkerInfo(ownerId_),
      PythonRRefFetchCall(ownerId_, rrefId()).toMessage(),
      true /* forceGradRecording */);

  const Message& message = futureResponse->wait();
  auto response = deserializeResponse(message);
  auto& rfr = unwrapAutogradMessage<PythonRRefFetchRet>(message, response);
  return PythonRpcHandler::getInstance().deserialize(
      SerializedPyObj::fromIValues(rfr.values()));
}

template class UserRRef<IValue>;
template class UserRRef<py::object>;

//////////////////////////  OwnerRRef  /////////////////////////////////////

template <typename T>
const T& OwnerRRef<T>::getValue() const {
  std::unique_lock<std::mutex> lock(mutex_);
  valueCV_.wait(lock, [this] { return value_.has_value(); });
  return value_.value();
}

template <typename T>
bool OwnerRRef<T>::hasValue() const {
  std::lock_guard<std::mutex> lock(mutex_);
  return value_.has_value();
}

template <typename T>
std::shared_ptr<FutureMessage> OwnerRRef<T>::getFuture() {
  std::unique_lock<std::mutex> lock(mutex_);
  if (future_.get()) {
    return future_;
  }
  future_ = std::make_shared<FutureMessage>();
  std::shared_ptr<FutureMessage> ret = future_;
  if (value_.has_value()) {
    lock.unlock();
    ret->markCompleted(Message());
  }
  return ret;
}

template <typename T>
void OwnerRRef<T>::setValue(T&& value) {
  std::unique_lock<std::mutex> lock(mutex_);
  value_ = std::move(value);
  std::shared_ptr<FutureMessage> future;
  future.swap(future_);
  lock.unlock();
  valueCV_.notify_all();
  if (future.get() && !future->completed()) {
    future->markCompleted(Message());
  }
}

template class OwnerRRef<IValue>;
template class OwnerRRef<py::object>;

} // namespace rpc
} // namespace distributed
} // namespace torch