// maximum number of in flight async nodes. Enqueuing of additional async ops
// after the limit is reached blocks until some inflight nodes finishes.
// The effect is bounding the memory held by inflight TensorHandles that are
// referenced by the inflight nodes.
// A recommended value has not been established.
// A value of 0 removes the limit, which is the behavior of TensorFlow 2.11.

    memcpy(buf, new_handle.c_str(), new_handle.size() + 1);
    *handle = buf;
  }
}

void TF_DeletePRunHandle(const char* handle) {
  delete[] handle;
  // TODO(suharshs): Free up any resources held by the partial run state.
}

void TF_SessionPRun(TF_Session* session, const char* handle,
                    const TF_Output* inputs, TF_Tensor* const* input_values,

  void UpdateLRU(const Key& key, const std::shared_ptr<Block>& block,
                 TF_Status* status) ABSL_LOCKS_EXCLUDED(mu_);

  /// Remove all blocks of a file, with mu_ already held.
  void RemoveFile_Locked(const std::string& filename)
      ABSL_EXCLUSIVE_LOCKS_REQUIRED(mu_);

  /// Remove the block `entry` from the block map and LRU list, and update the
  /// cache size accordingly.

  // TF_VariableInfo is constructed here by TensorFlow, and will be passed to
  // plugin as a opaque pointer. Plugin will need to call C APIs below to
  // operate on TF_VariableInfo (such as allocate temp tensor for the `var` held
  // by the underlying tensorflow::VariableInfo.
  TF_VariableInfo(int index, const std::string& name, tensorflow::Var* var) {
    var_info = tensorflow::VariableInfo{index, name, var};
  }

                                                  TF_Status* status);

// Creates a new operation - see `TF_NewOperation` for more details.
//
// The lock for `graph` must be held when calling this function.
//
// Unless implementing advanced behavior, like custom gradient functions, you
// most likely need to call `TF_NewOperation` instead.