Implicit conversion from float -100.94117647058823 to int loses precision