Implicit conversion from float -10.321543408360128 to int loses precision