Implicit conversion from float -9.625468164794015 to int loses precision