Implicit conversion from float -3.1190926275992297 to int loses precision