Implicit conversion from float -243.625 to int loses precision