Implicit conversion from float -12.589641434262958 to int loses precision