Implicit conversion from float -20.93525179856114 to int loses precision