Implicit conversion from float -54.402985074626855 to int loses precision