Implicit conversion from float -54.618834080717505 to int loses precision