Implicit conversion from float -0.9554140127388564 to int loses precision