Implicit conversion from float -9.438202247191 to int loses precision