Implicit conversion from float -0.9146341463414558 to int loses precision