Implicit conversion from float -3.16770186335404 to int loses precision