Implicit conversion from float -2.7966101694915295 to int loses precision