Implicit conversion from float -6.6763005780346845 to int loses precision