Implicit conversion from float -9.535315985130111 to int loses precision