Implicit conversion from float -1.030534351145036 to int loses precision