Implicit conversion from float -9.38811188811188 to int loses precision