Implicit conversion from float -12.21870047543581 to int loses precision