Implicit conversion from float -9.638989169675085 to int loses precision