Implicit conversion from float -7.735191637630663 to int loses precision