Implicit conversion from float -9.718309859154914 to int loses precision