TF_Status* status = TF_NewStatus();
  TF_NewWritableFile(full_path, &h, status);
  if (TF_GetCode(status) != TF_OK) {
    fprintf(stderr, "TF_NewWritableFile failed: %s\n", TF_Message(status));
    return 1;
  }
  fprintf(stderr, "wrote %s\n", full_path);
  free(full_path);
  TF_CloseWritableFile(h, status);
  if (TF_GetCode(status) != TF_OK) {
    fprintf(stderr, "TF_CloseWritableFile failed: %s\n", TF_Message(status));

#define TF_VALIDATE_STRUCT_SIZE(STRUCT_NAME, STRUCT_OBJ, SIZE_VALUE_NAME) \
  do {                                                                    \
    if (STRUCT_OBJ.struct_size == 0) {                                    \
      return tensorflow::Status(absl::StatusCode::kFailedPrecondition,    \
                                "Expected initialized `" #STRUCT_NAME     \

  // The value returned may be different from the one set by SetDeviceName, but
  // it will be compatible with it: the name will be updated by device placement
  // logic to refer to the specific device chosen.
  //
  // Example: If one calls `op->SetDeviceName("/device:GPU")`, the value
  // returned by DeviceName should be "/device:GPU:*" until a particular GPU is
  // chosen for the operation by the device placement logic in the

  // tasks to initialize the eager context and TF server for multi-client
  // execution.
  virtual Status EnableCollectiveOps(const ServerDef& server_def) = 0;

  // Check if the remote task is alive.
  virtual Status CheckRemoteAlive(const std::string& remote_task_name,
                                  bool* is_alive) = 0;

  // Get pointer to the coordination service agent instance.

// Set the shapes and types of the output's handle.
//
// The lengths of the arrays pointed to by `shapes`, `ranks`, and `types` must
// all be equal to `num_shapes_and_types`. If `ranks[i] != -1`, (i.e., if the
// rank is known), then it must be equal to the length of `shapes[i]`; if
// `ranks[i] == 1`, then `shapes[i]` may be nullptr.
//
// TODO(akshayka): Implement a corresponding getter method.

    }                                                             \
  } while (0)

#define TF_CLEANUP_AND_RETURN_IF_ERROR(C_STATUS, BUFFER, __VA_ARGS__) \
  do {                                                                \
    ::tensorflow::Status _s(__VA_ARGS__);                             \
    if (TF_PREDICT_TRUE(!_s.ok())) {                                  \
      TF_DeleteStatus(C_STATUS);                                      \

  //
  // For example some tensor handles may represent distributed values, in which
  // case placement information is lost when resolving the handle.
  //
  // If false, a caller might implement pretty-printing by resolving and
  // iterating over the resulting tensor. This may still be viable if resolving
  // the handle loses information, but `SummarizeValue` would be more precise.
  virtual bool PreferCustomSummarizer() const { return false; }

                                   absl::Span<AbstractTensorHandle*>)>;

// Runs `model` maybe wrapped in a function call op. This can be thought as
// being equivalent to the following python code.
//
// if use_function:
//   outputs = tf.function(model)(inputs)
// else:
//   outputs = model(inputs)
Status RunModel(Model model, AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,

class CGraphOptimizer : public CustomGraphOptimizer {
 public:
  explicit CGraphOptimizer(TP_Optimizer optimizer, const char* device_type)
      : optimizer_(optimizer), device_type_(device_type) {
    if (optimizer.create_func != nullptr) {
      c_optimizer_ = (*optimizer_.create_func)();
    } else {
      c_optimizer_ = nullptr;
    }
  }
  std::string name() const override { return "PluggableGraphOptimizer"; }

{
	int i, max;
	
	max = *(int*)p;
	for(i=0; i<max; i++)
		Add(i);
	return 0;
}

void
doAdd(int max, int nthread)
{
	enum { MaxThread = 20 };
	int i;
	uintptr_t thread_id[MaxThread];
	
	if(nthread > MaxThread)
		nthread = MaxThread;
	for(i=0; i<nthread; i++)
		thread_id[i] = _beginthreadex(0, 0, addThread, &max, 0, 0);
	for(i=0; i<nthread; i++) {
		WaitForSingleObject((HANDLE)thread_id[i], INFINITE);