Implicit conversion from float -1.4490861618799045 to int loses precision