Implicit conversion from float -12.873563218390814 to int loses precision