考虑以下C++代码实现的归并排序算法：

void merge(int arr[], int left, int mid, int right) {

      int n1 = mid - left + 1;

      int n2 = right - mid;

      int L[n1], R[n2];

      for (int i = 0; i < n1; i++)

            L[i] = arr[left + i];

      for (int j = 0; j < n2; j++)

            R[j] = arr[mid + 1 + j];

      int i = 0, j = 0, k = left;

      while (i < n1 && j < n2) {

            if (L[i] <= R[j]) {

                  arr[k] = L[i];

                  i++;

            }

            else {

                  arr[k] = R[j];

                  j++;

            }

            k++;

      }

      while (i < n1) {

            arr[k] = L[i];

            i++;

            k++;

      }

      while (j < n2) {

            arr[k] = R[j];

            j++;

            k++;

      }

}

void merge_sort(int arr[], int left, int right) {

      if (left < right) {

            int mid = left + (right - left) / 2;

            merge_sort(arr, left, mid);

            merge_sort(arr, mid + 1, right);

            merge(arr, left, mid, right);

      }

}

对长度为 n 的数组 arr ，挑用函数 merge_sort(a, 0, n-1) ，在排序过程中 merge 函数的递归调用次数大约是（    ）。