Implicit conversion from float -1.5730337078651644 to int loses precision