Implicit conversion from float -14.4222539229672 to int loses precision