Implicit conversion from float -5.547945205479451 to int loses precision