Implicit conversion from float -0.5336721728081244 to int loses precision