Implicit conversion from float -6.411856474258968 to int loses precision