seems i need new trunc implementation, any ideas how to make trunc faster?