Implicit conversion from float -91.10367892976586 to int loses precision